2
$\begingroup$

I am new to this data science stuff and I am trying a project on my own to learn more about this field. So I have a project that has the goal of taking in a bunch of features and indicating whether a player will make or miss a shot.

My current training data has a bunch of features alongside the output for each observation. I plan on using a Random Forrest model, as I am comfortable with it (and it fits the objective), however, one issue I see includes making sure luck does not play a role in the decision of the output.

I am trying to think of ways to limit the impact of luck on the model. For anyone familiar with basketball, sometimes a player takes a great shot and misses- sometimes he takes a horrible shots and makes it (both of those situat will be included in my training set). I do not want the model "thinking" that a shot is good/bad because of lucky/unlucky makes/misses.

So my question is how can I limit the impact of the luck within my data-sets or am I just able to assume that a large enough data set will take care of the luck since one gets lucky and unlucky at relatively equal rates (normal distribution) or do I instead revert to an unsupervised model that has the test data not include whether the shot was a miss or a make? Or is there another option to do something I have not considered to make the data better?

Thank you for your feedback.

$\endgroup$
1
$\begingroup$

If your training data is large enough, the model will have enough information to deal with chance through the statistics in the data. For example maybe a great shot is successful 80% of the time, so if there are 10 instances of great shot in the data there should be around 8 of them successful. In other words, the model will use the distribution of the data in order to make the best predictions possible.

When applying the model the predictions are deterministic, so there can be only one possible outcome for one instance. However with most types of models you can obtain the probability of success according to the model instead of a binary answer.

Minor notes:

  • lucky or unlucky would be a Bernouilli or binomial distribution, not a normal one.
  • unsupervised learning would be a completely different task, it would not make sense to do it for this reason.
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.